[R] SQL implementations (was: Are you experienced in SAS and R ...)
Prof Brian D Ripley
ripley at stats.ox.ac.uk
Sat Nov 24 09:47:05 CET 2001
On Fri, 23 Nov 2001, Dirk Eddelbuettel wrote:
> "Vadim" == Vadim Ogranovich <vograno at arbitrade.com> writes:
> Vadim> A related question. Prof. Ripley mentioned few times that one of his
> Vadim> approaches is to generate random samples from databases, which are
> Vadim> then analyzed in R (one doesn't need any sort of embedding for
> Vadim> this). Are there other techniques?
> You might be overlooking the fact that random subsampling from timeseries can
> severely distory the dynamic structure of the series.
Sure, but I've never seen a very long time series which was homogeneous
enough to justify being regarded as one population. For forecasting one
normally just uses the recent past.
Also, database theory (and some actual DBMSs) regard rows in tables as
unordered, hence the mathematical term `relation'. A `random sample' need
not be a *simple* random sample and does often need to be stratified
or clustered on important characteristics.
For some problems there are better large-data algorithms than those
currently implemented in R. But for others (general glms seems to be one),
there are none that do not require repeated access to the data. As Doug
and Duncan have pointed out, one could do some of this in the DBMS, and
some DBMSs may be a lot easier than others to do it in.
We do have a project to look at these ideas, but as any one takes a lot of
implementation it will take a while to yield results.
Brian D. Ripley, ripley at stats.ox.ac.uk
Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/
University of Oxford, Tel: +44 1865 272861 (self)
1 South Parks Road, +44 1865 272860 (secr)
Oxford OX1 3TG, UK Fax: +44 1865 272595
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !) To: r-help-request at stat.math.ethz.ch
More information about the R-help